How the robotic revolution will be a team effort

Ali Afrouzi, CEO, AI Incorporated 

Artificial intelligence is becoming more and more prevalent in our lives. In their current applications, artificial intelligence agents can learn and predict people’s behavior, helping them with useful suggestions and autonomously complete tasks assigned to them. At Silicon Valley’s AI Incorporated, they can now also form teams. “We have developed collaborative AI agents that are open to team-work and see other agents as potential collaborators,” says Ali Afrouzi, AI Incorporated’s CEO.

The artificial intelligence developed by AI Incorporated engineers works based on a reward system, which corresponds to a higher accuracy and efficiency in its performance. In environments crowded with traditional artificial intelligence agents, such as the internet, each artificial intelligence agent competes with other agents to maximise its own reward, thus maximising its own performance. In this example, a group of self-driving cars in a garage would compete with one another to exit first.

A competitive model does not consider the overall efficiency and outcome. Just as cases where in crowded human environments collaboration yields better results, machines would be better off if they collaborate with each other. This becomes exceedingly important if the machines are working on behalf their human owners, as is the case in the example of autonomous cars. AI Incorporated’s Collaborative Artificial Intelligence Technology (CAIT) is a new technological frontier that tackles this issue. “Our robots can work together to maximise their individual and collective performance through teamwork. They can form groups, share information, build tasks and even transfer their skills to one another,” says Afrouzi.

Some of the challenges AI Incorporated had to overcome were to enable CAIT agents to choose their teammates and deepen the relationship with the teammate that contributed the most. For this, CAIT agents must distinguish the degree of productivity among agents, giving higher productivity contributions more weight, and prune the ones that do not contribute. “We solved this issue using Hebbian Learning, a weight-variable describing the strength of each relationship which evolves based on the outcome of previous states and rewards,” says Afrouzi. “This allowed our systems to encourage CAIT agents to work together as a team towards a shared single goal, whereby the overall system attains more reward.” Individual agents also learn to bond with collaborators that contribute more. This also means that relationships with non-reciprocating agents can go sour. “In this way, our CAIT agents learn to see themselves and each other as a community of potential collaborators rather than lone agents competing against one another,” Afrouzi adds.

In addition to its patented Collective Artificial Intelligence Technology, AI Incorporated is the first company that works on Quantum SLAM in the field of mobile robotics. Simultaneous Localisation and Mapping, or SLAM, is a technology that allows robots to perceive their physical location in their surrounding environment. Quantum SLAM considers the complete state of the mechanical system at a given time, encoded as a phase point or a pure quantum state vector along with an equation of motion which carries the state forward in time.

Afrouzi is also the CEO of Bobsweep, a company specialising in cleaning robots which has sold hundreds of thousands of household robots since 2011. The company boasts six models in its portfolio, and plans to launch its most disruptive model in Jan 2019. “It is very AI rich, but its revolutionary features will be disclosed once the product is released,” said Afrouzi about Bobsweep’s upcoming product.

In explaining his vision for smart cities of future, Afrouzi envisions Passenger Pod as the system of transportation, which is a hybrid between owning an autonomous vehicle and using a ride-sharing system. “It uses the best of both worlds: pods are owned by passengers and are available in a variety of shapes and sizes. Unlike the other driverless cars, with this one you don’t have to worry about charging the battery, maintenance, or even a high price tag since the pod does not have locomotion components. The chassis is a separate interchangeable unit and can be shared among pod owners. Each chassis unit knows about the exact location of all the others and all the pods, which makes things so much easier.”

To start a conversation about the future of robotic technology visit


Get our latest features in your inbox

Join our community of business leaders